AI Agents · Operations · MCP2026
Chat-driven operations agents
MCP-connected agent swarm letting a non-technical owner run scheduling, dispatch, billing, and comms through natural-language chat.
5 → 1
dashboards replaced by one chat window
- Client
- Innovative Roofing Pros
- Year
- 2026
- Type
- AI Agents · Operations · MCP
What was the problem?
A roofing contractor was running the business through five disconnected tools: CRM, scheduling, dispatch, invoicing and comms. The owner and ops team weren't engineers. Every operational change meant clicking through each dashboard or asking someone technical to do it. Things slipped. Customers waited.
What did I build?
- Map the stack
- Expose through MCP
- Route requests
- One chat window
- 1Mapped the company's real stack and wrapped each operation as a tool: schedule, dispatch, invoice, look up.
- 2Exposed those tools to the agents through MCP (Model Context Protocol).
- 3Built a multi-agent layer that routes each request to the right tools across systems.
- 4Gave the owner and ops team one chat interface in place of the five dashboards.
What was the result?
- The owner runs operations conversationally instead of through five dashboards.
- Cross-tool coordination happens automatically.
- A non-technical team owns the system without engineering help.
What does it look like in practice?
- “Schedule a roof inspection for the Wilson job, Thursday morning, send confirmation.”
- “Show me everyone whose deposit hasn't cleared this week.”
- “Reassign Tuesday's installs — Mike called in sick.”
What is it built with?
- MCP (Model Context Protocol)
- Multi-agent architecture
- Custom tool wrappers
- Anthropic / OpenAI
- ai agents
- mcp
- operations
- non-technical